Wolfgang >The correction comes from the fact that Christian's formula for OR >predicates is correct for predicates involving different columns. >Expressing it in probability terms: > >p( A or B ) = p(A) + P(B) - p(A and B) > >In this case the optimizer knows that the two sides of the OR can not >both be true, i.e. p(A and B) = selectivity1*selectivity2 = 0 I stand corrected. Thank you. I should probably avoid replying to email after 2am ;-) Cheers, Chris -- //www.freelists.org/webpage/oracle-l